Thread: Shouldn't we document "don't use a mountpoint as $PGDATA"?
I had a discussion with some folks at Red Hat about this: https://bugzilla.redhat.com/show_bug.cgi?id=1247477 I had the idea that we had documented somewhere that the data directory should not be a filesystem mount point, but I sure can't find it now. Any objections to adding some text about this to section 17.2, near the caveats about NFS? regards, tom lane
On 07/28/2015 11:01 AM, Tom Lane wrote: > I had a discussion with some folks at Red Hat about this: > https://bugzilla.redhat.com/show_bug.cgi?id=1247477 > > I had the idea that we had documented somewhere that the data directory > should not be a filesystem mount point, but I sure can't find it now. > Any objections to adding some text about this to section 17.2, near the > caveats about NFS? Nope. This is one of those things which seems obvious, but isn't to newcomers who aren't former sysadmins. -- Josh Berkus PostgreSQL Experts Inc. http://pgexperts.com
On 07/28/2015 02:01 PM, Tom Lane wrote: > I had a discussion with some folks at Red Hat about this: > https://bugzilla.redhat.com/show_bug.cgi?id=1247477 > > I had the idea that we had documented somewhere that the data directory > should not be a filesystem mount point, but I sure can't find it now. > Any objections to adding some text about this to section 17.2, near the > caveats about NFS? Please do. I came across a client who should have known better doing this just a week or two ago. cheers andrew
On Tue, Jul 28, 2015 at 02:19:00PM -0400, Andrew Dunstan wrote: > > On 07/28/2015 02:01 PM, Tom Lane wrote: > >I had a discussion with some folks at Red Hat about this: > >https://bugzilla.redhat.com/show_bug.cgi?id=1247477 > > > >I had the idea that we had documented somewhere that the data directory > >should not be a filesystem mount point, but I sure can't find it now. > >Any objections to adding some text about this to section 17.2, near the > >caveats about NFS? > > > > Please do. I came across a client who should have known better doing > this just a week or two ago. Uh, we added an initdb warning about mount points: commit 17f15239325a88581bb4f9cf91d38005f1f52d69Author: Bruce Momjian <bruce@momjian.us>Date: Sat Feb 16 18:52:50 2013 -0500 Warn about initdb using mount-points Add code to detect and warn about trying to initdb or create pg_xlog on mount points. Was that not sufficient? -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. +
On Tue, Sep 1, 2015 at 08:20:47PM -0400, Bruce Momjian wrote: > On Tue, Jul 28, 2015 at 02:19:00PM -0400, Andrew Dunstan wrote: > > > > On 07/28/2015 02:01 PM, Tom Lane wrote: > > >I had a discussion with some folks at Red Hat about this: > > >https://bugzilla.redhat.com/show_bug.cgi?id=1247477 > > > > > >I had the idea that we had documented somewhere that the data directory > > >should not be a filesystem mount point, but I sure can't find it now. > > >Any objections to adding some text about this to section 17.2, near the > > >caveats about NFS? > > > > > > > > Please do. I came across a client who should have known better doing > > this just a week or two ago. > > Uh, we added an initdb warning about mount points: > > commit 17f15239325a88581bb4f9cf91d38005f1f52d69 > Author: Bruce Momjian <bruce@momjian.us> > Date: Sat Feb 16 18:52:50 2013 -0500 > > Warn about initdb using mount-points > > Add code to detect and warn about trying to initdb or create pg_xlog on > mount points. > > Was that not sufficient? Oh, I see in the commit message you knew about the initdb warning --- never mind. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. +